Skip to main content

Resultados relacionados

Resultados totales (Incluyendo duplicados): 2
Encontrada(s) 1 página(s)

Minerva. Repositorio Institucional de la Universidad de Santiago de Compostela
Publication Capítulo o parte de un libro (BookPart)

Consistency of Explanations in Evolving Supervised Learning Contexts

  • Ostrowski, Marcin
  • Kaczmarek-Majer, Katarzyna
  • Alonso Moral, José María
In many real-world scenarios, explanations are not perceived as isolated entities, but rather as streams accompanying models that operate over extended periods of time. In such contexts, users are exposed not to a single explanation, but to a sequence of explanations whose evolution may influence trust, understanding, and decision-making. In this work, we aim to capture the evolution of explanations that exhibit sudden changes, contradict prior explanations, or demonstrate erratic fluctuations, as these may compromise their practical utility, even when maintaining a high level of predictive performance. We propose a framework aimed at supporting the assessment of the consistency of explanations for evolving supervised learning contexts, taking into account the variability of the input datasets, and the uncertainty of the model. We introduce a model-agnostic and explanation-agnostic index for quantifying the temporal consistency of automated explanations. We empirically demonstrate that the proposed index facilitates the quantification and localization of temporal instability in explanation streams. Experimental results for simulated multivariate time series underscores the ability of the proposed index to capture instantaneous changes and their relative impact in relation to historical patterns.
RUC. Repositorio da Universidade da Coruña
Publication Artículo científico (JournalArticle)

Automatic Criteria for Prioritizing Software Requirements in Spanish Projects

  • Limaylla-Lunarejo, María-Isabel
  • Condori Fernández, Nelly
  • Rodríguez Luaces, Miguel
Financiado para publicación en acceso aberto: Universidade da Coruña/CISUG The final version of the PrioReSpa dataset, including requirement descriptions, relevance label, F/NF classification, MoSCoW categorization, and Mandatory criteria, has been published on Zenodo. https://doi.org/10.5281/zenodo.15552978, [Abstract]: Requirements prioritization is a complex process for determining the implementation order of requirements based on business value, cost, time, and other factors. Despite the growing use of AI to automate prioritization, full automation remains limited because common criteria, such as stakeholder ranking, cost, and dependencies, still require human input. To mitigate this, we focus on criteria that can be obtained automatically. We identified three such criteria: (1) Functional/Non-Functional (F/NF) classification, based on requirement purpose using pre-trained models; (2) MoSCoW categorization, which uses linguistic patterns and keywords to assess necessity; and (3) a Mandatory criterion derived from the similarity requirements process. In this study, we investigate the prioritization of requirements written in Spanish using automated criteria and automating the process with the LambdaMART algorithm. Additionally, we introduce PrioReSpa, a dataset of 401 requirements collected from a GIS project based on a software product line, each labeled with an importance score and the three prioritization criteria. Our findings demonstrate the effectiveness of using automated criteria for requirements prioritization. Results show that F/NF classification is the most influential criterion, followed by MoSCoW, while the Mandatory criterion had no impact. We trained LambdaMART models using XGBRanker and LGBMRanker implementations, performing hyperparameter optimization with Optuna. Both rankers obtained comparable NDCG performance, with LightGBM slightly outperforming in ranking metrics and XGBoost providing faster training. These findings demonstrate the viability of using LambdaMART algorithm to generate initial ranked lists of requirements, particularly for the top 10–20 priorities, reducing stakeholder involvement in early stages., This research was partially supported by the following grants: “NextGenerationEU”/PRTR; PID2022-141027NB-C21 (EarthDL), partially funded by MCIN/AEI/ 10.13039/ 501100011033 and the EU/ERDF, “A way of making Europe´´ (1st and 3rd authors); and 101134894-HORIZON-CL6-2023-GOVERNANCE-01 funded by: European Union´s Horizon Europe research and innovation programme, and ED431G-2023/04 funded by the Xunta de Galicia - Conselleria de Educación, Ciencia, Universidades e Formación and the EU/ERDF (2nd author). Funding for open access charge: Universidade da Coruña/CISUG., Xunta de Galicia; ED431G-2023/04